Machine-made sand grading calculation method based on double-vision compensation algorithm

Abstract

The invention discloses a machine-made sand grading calculation method based on a double-vision compensation algorithm, and relates to the technical field of computer vision. According to the invention, through a compensation algorithm combining global and local images, the stability problem of small-particle-size particles is effectively solved. The global image provides distribution information of the whole sand grains, the local image can accurately capture details of the small grains, the instability of the small-grain-size grains is compensated through an algorithm, and the stability and the accuracy of a measurement result are ensured; according to the method, errors caused by particle shape irregularity are overcome, the mass of the particles is measured through experiments, the equivalent height is calculated by combining the projected area and density of the particles, and the method effectively avoids volume calculation deviation caused by the particle shape irregularity.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and specifically to a method for calculating the gradation of manufactured sand based on a dual vision compensation algorithm. Background Art

[0002] The accurate measurement of sand grain gradation is crucial for industrial fields such as building materials and mineral processing, because the size distribution of particles directly affects the performance and processing of materials, especially in the quality control of construction sand. A reasonable sand grain gradation can effectively reduce the accumulation pores in concrete, thereby reducing the amount of cement slurry used, improving the quality of concrete and reducing costs. However, the traditional screening method has disadvantages such as large workload and strong subjectivity, and may damage the original particle size distribution of sand grains during the vibration screening process, especially the measurement of fine particles is particularly difficult [1].

[0003] In order to overcome the defects of the traditional screening method, a gradation detection method based on the image method has emerged. The image method includes the static image method and the dynamic image method. The static image method is often affected by particle adhesion and overlap, resulting in low gradation detection accuracy. Although the dynamic image method can avoid the defects of the static image method, it has high requirements for industrial cameras and is difficult to capture fine sand grains with a particle size less than 0.15 mm [2][3]. In existing inventions, the dual-camera multi-scale method proposed by Wen Hua Lin and the gradation compensation model of Xiao Yu Huang have made remarkable progress in sand grain analysis. Especially, the multi-scale method can effectively analyze in different particle size ranges [4][5], but there are still certain limitations in dealing with fine particles.

[0004] For the measurement of fine particles, the sand grain classification algorithm based on image processing proposed by Zhang Ming Li provides a new idea for particle size distribution analysis. However, the recognition accuracy of fine particles still needs to be further optimized. [6] Although Li XiaoHong's dynamic light scattering technology has advantages in the measurement of nano-scale particle sizes, it faces certain challenges when dealing with larger particle size sand grains. [7] In addition, Wang Da Wei's acoustic particle size analysis method performs excellently in the rapid measurement of large particle size sand grains, but the resolution of fine particles still needs to be improved [8].

[0005] Although the existing inventions have made remarkable progress in sand grain analysis, the accurate measurement of fine particles still faces challenges.

[0006] [1] Sha Aimin, Wang Chaofan, and Sun Chaoyun, A method for detecting the gradation of mineral aggregates in asphalt mixtures based on images [J]. Journal of Chang'an University (Natural Science Edition). 2010. 30(05): 1-5.

[0007] [2] Wang Xiaoyue, Invention on the Influence Law of Particle Shape on the Numerical Simulation of Vibration Screens. 2017.

[0008] [3] Zhou Jianhua, Fang Huaiying, Yang Jianhong, et al. Selection and Experiment Invention of Aggregate Particle Size Detection and Characterization Parameters by Image Method [J]. Acta Metrologica Sinica, 2018, 39(6): 783-790. DOI: 10.3969 / j.issn.1000-1158.2018.06.06.

[0009] [4] Lin Wenhua, Fang Huaiying, Fan Lulu, Yang Jianhong. Measurement of Mechanism Sand Gradation and Prediction of Void Ratio by Dual-Camera Multi-Scale Method [J]. Journal of Huaqiao University (Natural Science Edition), 2022, 43(03): 285-290.

[0010] [5] Huang Xiaoyu. Invention on the Measurement of Mechanism Sand Gradation and Compensation Algorithm Based on Dynamic Image Method [D]. Huaqiao University, 2020.

[0011] [6] Zhang MingLi, "Image Processing-Based Sand Particle Classification Algorithm," IEEE Transactions on Geoscience and Remote Sensing, vol. 57, no. 10, pp. 7892-7903, 2019.

[0012] [7] Li XiaoHong, "Dynamic Light Scattering Technique for Nanoscale Particle Size Measurement," Journal of Colloid and Interface Science, vol. 550, pp. 1-9, 2019.

[0013] [8] Wang DaWei, "Acoustic Particle Size Analysis for Rapid Measurement of Large Sand Particles," Journal of Applied Geophysics, vol. 168, pp. 103801, 2019.

[0014] Therefore, a new solution needs to be proposed for the above problems. Summary of the Invention

[0015] The object of the present invention is to provide a calculation method for the gradation of manufactured sand based on a dual-vision compensation algorithm, which compensates for the measurement accuracy of sand grains in different particle size ranges by combining global and local image data, especially for effective compensation in the analysis of particles smaller than a certain particle size threshold (such as 0.15 mm or other particle sizes). By introducing the concept of "particle size threshold" and combining the different field of view and resolution data of the global camera and the local camera, the distribution of fine particles in the global image is deduced, thereby improving the measurement accuracy of small particles and solving the technical problems proposed in the background art.

[0016] To achieve the above object, the present invention provides the following technical solution: A calculation method for the gradation of manufactured sand based on a dual-vision compensation algorithm, at least including the following steps:

[0017] S1: Use a dual camera to collect the global image and local image of the sand, and use the multiple vibration shooting sampling technique to obtain sand grain pictures. For the different distribution states of the same group of sand grains during the vibration of the vibrating disk, multiple shooting samplings are carried out. Each time of shooting, the distribution and arrangement state of the sand grains will change, thus effectively reducing the errors that may occur in a single shooting.

[0018] S2: Preprocess the taken sand grain pictures. Since the sand grain information to be extracted from the global image and the local image is different, two image preprocessing methods are adopted. The global image processing focuses on stability and overall contour information, while the local image processing focuses on high precision and detail retention to avoid large error precision problems in subsequent global gradation calculations.

[0019] S3: After extracting the sand grain contour information after image preprocessing, it is necessary to convert the actual sand grain size into the pixel size in the image to accurately divide and classify the particle size ranges in the image.

[0020] S4: After dividing the particle sizes of all sand grains, it is necessary to calculate the contour projection area of each sand grain for use in the subsequent calculation of the projection area ratio of the particle size range.

[0021] S5: After obtaining the contour projection areas of all sand grains, calculate the total projection area of the sand grains in the same particle size range and calculate the projection area ratio of the corresponding particle size range larger than.

[0022] S6: After calculating the projection area ratios of each particle size range larger than the particle size threshold, use three methods: the adjacent particle size range method, the average value method, and the weighted average method to calculate the projection area ratios of the small particle size sand grains in the global image and the local image, that is, the projection area ratios of the small particle size ranges. This ratio can be used to obtain the total projection area of the small particle size sand grains in the global image, where D tis the particle size threshold, and R is the projected area ratio. After calculating the projected area ratios of the small particle size intervals deduced by the adjacent particle size interval method, the average value method, and the weighted average method, the variance of each method is calculated respectively. The variance reflects the consistency and stability of the deduced results of each method. The smaller the variance, the more stable the result. By comparing the variances of these three methods, the method with the smallest variance is selected as the final calculation method to ensure that the deduced result of the projected area ratio of the small particle size interval has the lowest error and the best stability, which can improve the overall calculation accuracy and ensure the reliability of the measurement results;

[0023] S7: Obtain the projected area ratio of the small particle size interval Map the small-sized sand grains in the local image to the global image to improve the area information of the small-sized sand grains in the global image. Then, combined with the height H obtained in the single gradation shooting experiment, calculate the volume of the sand grains in all particle size intervals in the global image, and thus calculate the gradation result of this group of sand grains.

[0024] Furthermore, the preprocessing of the global image in S2 at least includes the following steps:

[0025] First, perform grayscale conversion and Gaussian filtering for denoising to retain the main contours. The standard deviation of the Gaussian filtering is 2;

[0026] Adopt morphological "top-hat transformation" to highlight the bright areas, combine distance transformation to mark the particle centers, and cut the adhered particles through the watershed algorithm;

[0027] Finally, perform Otsu adaptive binarization to extract clear contours.

[0028] Furthermore, the preprocessing of the local image in S2 at least includes the following steps:

[0029] After grayscale processing, use mild Gaussian filtering (standard deviation is 1) to retain details, and the standard deviation is 1;

[0030] Perform morphological "closing operation" to fill small gaps, combine distance transformation and watershed algorithm to segment adhered particles;

[0031] Use Canny edge detection to extract particle contours, and finally use Otsu adaptive binarization to enhance details. The threshold of the Canny edge detection is 50 - 150.

[0032] Furthermore, during the conversion in S3, due to different camera parameters (such as focal length, object distance) and image resolution, the actual pixel length of the sand grain size displayed in the image is different,

[0033] Therefore, use the pixel conversion formula to convert the actual particle size (in millimeters) to the corresponding pixel length in the image:

[0034]

[0035] Among them, P represents the pixel value, f represents the camera focal length, fd represents the camera object distance, and D real represents the true size of the object.

[0036] Furthermore, the contour projection area calculation method used to calculate the contour projection area of each sand grain in S4 is the method of expanding half a pixel point outward at 45 degrees. This contour projection area calculation method effectively compensates for the influence of pixel discreteness and resolution on area estimation, thereby improving the accuracy of projection area calculation;

[0037] The specific operation of the contour projection area calculation method is to expand the contour of each sand grain along the 45-degree diagonal direction and calculate the pixel area occupied by the expanded contour, that is, to count the number of pixels in the filled area of the expanded contour, that is, the number of pixels in the white area, and use this as the contour projection area of the sand grain.

[0038] Furthermore, the steps of S5 are described as follows:

[0039] After obtaining the contour projection areas of all sand grains, calculate the total projection area of sand grains in the same particle size range and calculate the projection area ratio of the corresponding particle size range with a particle size greater than D t where

[0040]

[0041] Among them, is the total projection area of sand grains in a certain particle size range [D i , D i+1 in the global image,

[0042] is the total projection area of sand grains in the corresponding particle size range in the local image.

[0043] Among them, the sand grain projection area in the global image in the projection area ratio refers to the sand grain projection area in the entire global image in this particle size range, and does not separately remove the sand grain contour information in the local area.

[0044] Furthermore, the adjacent particle size range method in S6 is applicable to the case where the particle distribution is relatively uniform. Among them, the projection area ratio of adjacent particle size ranges is usually relatively stable. Refer to the following formula:

[0045]

[0046] Among them: the particle size threshold is D t , select two particle size ranges [D t , Dt+1 and [D t+1 , D t+2 , and the projected area ratio in the small particle size range is estimated by the average value of their projected area ratios.

[0047] Furthermore, the average value method in S6 is applicable to the case where the particle size distribution is relatively uniform. Because it assumes that the change in the projected area ratio of the particle size range is relatively smooth and there is no obvious uneven distribution, the projected area ratio in the small particle size range is estimated by calculating the average value of the projected area ratios of all known particle size ranges;

[0048] First, calculate the projected area ratio of each particle size range, and then average the ratios of the projected area ratios to obtain the average projected area ratio

[0049] Next, select the two particle size ranges closest to the above average value, and estimate the projected area ratio in the small particle size range through the projected area ratios of these two ranges. Refer to the following formula:

[0050]

[0051] where n is the total number of particle size ranges, t represents the position corresponding to the particle size threshold, which is used to determine the threshold division point of the particle size range, so that the ranges after t are regarded as ranges larger than the particle size threshold.

[0052] Then, according to the average projected area ratio, select the two projected area ratios closest to it. At this time, the projected area ratio in the small particle size range is calculated from the average value of these two projected area ratios.

[0053] Furthermore, the weighted average method in S6 calculates the projected area ratio in the small particle size range by setting weighted coefficients according to the distribution of sand grains in different particle size ranges. The weighted average method assumes that ranges with a larger number of particles should be given a higher weight, while ranges with a smaller number of particles should be given a lower weight. The weighted coefficients are usually set based on the number and frequency distribution of particles in the particle size range. Through weighted averaging, a more accurate projected area ratio can be obtained, especially applicable to sand grains with uneven particle size distribution. The formula is as follows:

[0054]

[0055] where represents the projected area ratio of the i-th particle size range larger than the particle size range, n is the total number of particle size ranges, t represents the position corresponding to the particle size threshold, which is used to determine the threshold division point of the particle size range, so that the ranges after t are regarded as ranges larger than the particle size threshold, α iis the corresponding weight coefficient, which is determined according to the number and frequency of sand particles in each particle size interval. The particle size interval with a larger number of particles corresponds to a higher weight coefficient. i Need to meet:

[0056]

[0057] Further, the S7 at least includes the following steps:

[0058] The total projected area of ​​small-size sand particles in the global image is defined as Total projected area of ​​small-size sand particles in the local image

[0059] The total projected area of ​​sand particles in a certain size range in the global image is The total projected area of ​​sand particles in a certain size range in the local image is The projection area ratio is R, and the projection area ratio smaller than the particle size threshold is The projected area ratio of a certain particle size interval greater than the particle size threshold is R i , where i∈{t,…,n-1}, t represents the position corresponding to the particle size threshold, which is used to determine the threshold dividing point of the particle size interval;

[0060] The total projected area of ​​small-size sand particles in the local image and projected area ratio The total projected area of ​​small-size sand particles in the global image is obtained by multiplication:

[0061]

[0062] is the particle size in the global image that is smaller than the particle size threshold D t The total projected area of ​​the sand particles. It should be noted that as long as the particle size is smaller than the threshold D t When calculating the total projected area of ​​these sand particles in the global image, the compensation coefficients are unified as

[0063] Based on the fixed assumption of sand and gravel density ρ, using the formula Calculate the equivalent height H of specific graded particles;

[0064] The mass M of sand and gravel was determined experimentally, and pictures of multiple groups of single-graded (i.e., single-size interval) sand grains were taken to calculate the total projected area S of each group of sand grains, and the formula Calculate the equivalent height h of each group of sand particles, then take the average of multiple groups of h to get the equivalent height H;

[0065] The total projected area of ​​each group of sand particles is calculated by the forty-five degree outward expansion method of half a pixel point.

[0066] Finally, a new method for characterizing the equivalent volume of sand and gravel is proposed:

[0067] Projected area of sand grains × equivalent height H

[0068] The object of the new method for characterizing the equivalent volume of sand and gravel is the same batch of materials, that is, all sand grains with the same density, which is used to calculate the volume of each sand grain. The sand grain is a sand grain taken by normal sampling and shooting. By combining the projected area of the particle with the deduced equivalent height, the error caused by the irregularity of the particle shape is effectively overcome, and the accuracy of volume calculation is significantly improved.

[0069] Compared with the prior art, the beneficial effects of the present invention are:

[0070] 1. The present invention effectively solves the stability problem of small-particle-size particles by combining the compensation algorithms of global and local images. The global image provides the distribution information of the overall sand grains, while the local image can accurately capture the details of small particles. By compensating the instability of small-particle-size particles through the algorithm, the stability and accuracy of the measurement results are ensured;

[0071] 2. The present invention overcomes the error caused by the irregularity of particle morphology. Traditional volume calculation methods assume that the particle morphology is regular, and usually estimate the volume of particles by multiplying the projected area by the assumed height. However, the actual shape of particles is usually irregular, and the actual height of particles cannot be directly obtained from two-dimensional images, resulting in a large volume calculation error. The present invention determines the particle mass through experiments, and combines the projected area and density of the particles to deduce an equivalent height, which effectively avoids the volume calculation deviation caused by the irregular particle shape;

[0072] 3. The present invention solves the problem of data fusion between global and local images. During multiple shootings and vibrations, the inconsistency of particle distribution in global and local images poses a challenge to data fusion. The present invention overcomes the fusion problem caused by the distribution difference between global and local images by adopting a compensation algorithm. Through precise ratio calculation and compensation algorithm, the small-particle information in the local image can be organically fused with the large-particle data in the global image, thus realizing the effective integration of global and local image data and ensuring the comprehensiveness and consistency of particle gradation calculation;

[0073] 4. The present invention adopts the multiple vibration shooting and sampling technology. Under the action of the vibrating disk, the same group of sand grains is photographed and sampled multiple times in different distribution states to ensure the diversity of particle distribution, thereby effectively reducing the error caused by uneven distribution in a single shooting. By averaging the mass ratio of each particle size interval in different states, the gradation results of multiple shootings are integrated into an overall result, which can effectively correct the error and ensure the accuracy of the final calculation result. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0075] Figure 1 is a structural schematic diagram of the present invention;

[0076] Figure 2 is a flowchart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0077] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments.

[0078] The purpose of the present invention is to solve the problems of insufficient accuracy and sampling limitations in the existing sand particle size grading measurement methods, especially in the measurement of fine particles. In the traditional single-camera system, when capturing particles smaller than a certain threshold (for example, 0.15 mm), due to resolution limitations and shooting instability, accurate data of the particles cannot be effectively obtained, thus affecting the accuracy of grading calculation. To overcome these problems, the present invention proposes a solution based on a dual-camera system. Through the collaborative work of the global camera and the local camera, combined with the particle size threshold, projection area ratio, and equivalent volume characterization method, accurate measurement of fine particles and grading calculation are achieved. Through this innovative method, the accuracy and stability of particle size grading measurement can be effectively improved, especially in the case of uneven distribution of fine particles and irregular particle shapes, ensuring the reasonable integration of global and local data, thereby providing more reliable grading calculation results, and being widely applicable to fields such as building materials and mineral processing.

[0079] In the invention, in order to effectively measure the mass content of sand particles, it is a feasible method to use the volume ratio instead of the mass ratio. This is based on the assumption that the density ρ of the same batch of sand particles is approximately equal. The calculation formula for the mass content h is:

[0080]

[0081] where m is the mass of sand particles in a certain interval, M is the total mass of sand particles, V i is the volume of each sand particle, V j is the total volume of sand particles. Since the density is approximately equal, the mass ratio can be approximately replaced by the volume ratio approximately.

[0082] In terms of particle volume calculation, the present invention proposes a new equivalent volume characterization method. Since the morphology of manufactured sand particles is irregular and the captured images are two-dimensional, it is impossible to directly obtain the actual height of the particles, and the traditional volume calculation method (i.e., area multiplied by height) is difficult to apply. To solve this problem, the present invention calculates by means of probability statistics: take pictures of multiple groups of single-sized (i.e., a single particle size range) sand grains, experimentally measure the mass of the sand and gravel as M, calculate the total projected area S of each group of sand grains, and use the formula to deduce the equivalent height h of each group of sand grains. Then, take the average value of multiple groups of h to obtain the equivalent height H. Finally, a new equivalent volume characterization method for sand and gravel is proposed: the projected area of sand grains × the equivalent height H; by combining the projected area of the particles with the deduced equivalent height, the error caused by the irregularity of the particle shape is effectively overcome, and the accuracy of volume calculation is significantly improved.

[0083] To better solve the limitations of global images and local images in particle sampling, the present invention introduces the concept of particle size threshold D t . This parameter is determined by the global sampling camera and represents the smallest particle size that the global camera can reliably capture. When setting the particle size threshold, if the threshold is within a certain particle size range, the particle size threshold will take the upper limit value of that range. For example, if the particle size threshold falls within the range of [D i , D i+1 , then the particle size threshold takes D i+1 . This rounding method ensures that the particle size threshold always corresponds to the upper limit of the particle size range, facilitating subsequent calculations and deductions.

[0084] For particles larger than this threshold, although the local camera can effectively capture images, due to the small number of large particles, the number that the local camera can capture is limited. As a result, when estimating the sand grain gradation through the local camera, the reliability of the gradation of larger particles is relatively low. Therefore, it is necessary to combine the data of global and local images. The global camera captures images of all sand grains, mainly for collecting images of sand grains with larger particle sizes; due to the small object distance, the local camera can only capture images of local sand grains in the vibrating tray, but can collect images of sand grains with smaller particle sizes. More accurate and stable gradation calculation is achieved by comprehensively analyzing the data of large particle sizes and small particle sizes.

[0085] The difficulty in using dual cameras to detect sand gradation is that the global camera cannot capture images of small particle size grains, and the local camera collects particle images of some areas in a sampling manner and cannot directly integrate the collected data into the data of the global camera to calculate the overall sand grain gradation.

[0086] The core problem of the present invention is how to utilize the particle size threshold D t, combining the data of the global and local images, to deduce the distribution of particles smaller than the threshold in the global image. To achieve this goal, we propose a compensation algorithm based on the projected area ratio.

[0087] The present invention defines the total projected area of small-sized sand grains in the global image as The total projected area of small-sized sand grains in the local image The total projected area of sand grains in a certain particle size range in the global image is The total projected area of sand grains in a certain particle size range in the local image is The projected area ratio is R, and the projected area ratio of particles smaller than the particle size threshold is The projected area ratio of a certain particle size range larger than the particle size threshold is R i . (Where i ∈ {t, …, n - 1}, t represents the position corresponding to the particle size threshold, which is used to determine the threshold division point of the particle size range).

[0088] Through the total projected area of small-sized sand grains in the local image and the projected area ratio The product is used to obtain the total projected area of small-sized sand grains in the global image:

[0089]

[0090] is the total projected area of sand grains smaller than the particle size threshold D t in the global image. It should be noted that as long as the particle size is smaller than the threshold D t , when calculating the total projected area of these sand grains in the global image, the compensation coefficient is uniformly

[0091] When calculating the projected area ratio, although the contour area and short diameter value calculated in different images will be affected by factors such as shooting conditions and resolution, these effects are fixed for the same group of sand grains. Therefore, the present invention chooses to uniformly incorporate these factors into the calculation of the projected area ratio without additional conversion steps. In this way, the original contour parameters can be directly used to calculate the projected area ratio, avoiding the complexity brought by conversion and ensuring the accuracy of the calculation of the global small particle size projected area.

[0092] The projected area ratio R is a key parameter we proposed, which is used to compensate for the resolution difference between the global image and the local image and deduce the projected area of particles smaller than the particle size threshold D t in the global image. The global image can clearly capture each particle larger than the particle size threshold. Therefore, for each particle size range [D i , D i+1 (where D i > Dt ), and its corresponding projected area ratio can be directly calculated from the global image.

[0093] However, for particles smaller than the particle size threshold D t , since the projected area data of these particles cannot be directly obtained from the global image, it is necessary to compensate through the information of small particles in the local image. The local image focuses on photographing small particles and can provide high-resolution detail data. By comparing the common contour information of small particles in the local image with large particles in the global image, we propose three calculation methods to calculate the projected area ratio of small-sized particles in the global image using a compensation algorithm so as to make up for the lack of information of particles smaller than the threshold in the global image. The three projected area ratios of small particle sizes calculation methods:

[0094] 1. Adjacent particle size interval method: Select two particle size intervals close to the particle size threshold and calculate the projected area ratio of the small particle size interval through the ratio of their projected areas. This method is applicable to the case of relatively uniform particle size distribution.

[0095] 2. Average value method: Calculate the average value of the projected area ratios of all known particle size intervals, and select the ratio of the two intervals closest to the average value as the projected area ratio of the small particle size interval.

[0096] 3. Weighted average method: According to the sand particle distribution, set the weighting coefficient, calculate the weighted average value of different particle size intervals, and obtain a more accurate projected area ratio. This method is applicable to the case of non-uniform particle size distribution.

[0097] The calculated by these methods can effectively compensate the projected area information of small particles in the local image into the global image, ensuring the accurate restoration of the particle gradation in all particle size intervals in the global image. Even in the case of non-uniform particle distribution, this compensation algorithm can accurately restore the small particle gradation in the global image, thus ensuring the accuracy of the final gradation calculation.

[0098] The first innovation point is that the present invention proposes a new equivalent volume characterization method, which solves the error problem in traditional volume calculation caused by irregular particle morphology and two-dimensional images. By experimentally measuring the sand particle mass M and projected area S, and combining with the particle density ρ, use the formula to calculate the equivalent height H of the particle. This method combines the projected area with the calculated equivalent height to form a new volume calculation method projected area S × height H, effectively overcoming the error caused by irregular morphology and improving the volume calculation accuracy.

[0099] The second innovation point lies in combining image preprocessing and a dual-vision compensation algorithm to achieve high-precision calculation of sand particle gradation. First, through image preprocessing technology, the contour information of sand particles is extracted, and considering the shooting limitations of the global camera (such as resolution limitations and the minimum recognizable particle size), a particle size threshold is determined to divide the sand particles into particle size intervals. On this basis, the "45-degree outward expansion by half a pixel method" is used to accurately calculate the projected area of sand particles, reducing the area error caused by pixel discontinuity. Then, by calculating the ratio of the projected areas of each particle size interval, further combining the particle size information in the global image and the local image, three different methods (adjacent particle size interval method, average value method, and weighted average method) are used to calculate the ratio of the projected areas of small particle size sand particles respectively. This ratio is used to estimate the total projected area of small particle size sand particles in the global image. Finally, combining the estimated projected area ratio and the equivalent height H of the particles, the present invention calculates the volume of sand particles in each particle size interval and finally obtains the gradation and fineness modulus difference of sand particles. Through this high-precision algorithm process, the particle size distribution of sand particles can be more accurately characterized, thereby optimizing the gradation calculation of sand and gravel and improving the accuracy and reliability of sand and gravel quality control.

[0100] Example 1:

[0101] Based on the above description, a mechanism sand gradation calculation method based on a dual-vision compensation algorithm is summarized in this case;

[0102] Please refer to Figure 1 and Figure 2 , a mechanism sand gradation calculation method based on a dual-vision compensation algorithm, at least includes the following steps:

[0103] S1: Use dual cameras to collect the global image and local image of sand, and use the multiple vibration shooting sampling technology to obtain sand particle pictures. For the different distribution states of the same group of sand particles during the vibration of the vibrating disk, multiple shooting samplings are carried out. Each time of shooting, the distribution and arrangement states of sand particles will change, thus effectively reducing the errors that may occur in a single shooting, such as particle adhesion and uneven distribution, etc.;

[0104] S2: Preprocess the taken sand particle pictures. Since the sand particle information to be extracted for the global image and the local image is different, two image preprocessing methods are adopted. The global image processing focuses on stability and overall contour information, while the local image processing focuses on high precision and detail retention to avoid large error precision problems in subsequent global gradation calculations;

[0105] S3: After extracting the contour information of sand particles after image preprocessing, it is necessary to convert the actual sand particle size into the pixel size in the image to accurately divide and classify the particle size intervals in the image;

[0106] S4: After classifying all sand grains by particle size, it is necessary to calculate the contour projection area of each sand grain for use in subsequent calculations of the projection area ratio for each particle size range.

[0107] S5: After obtaining the contour projection areas of all sand grains, calculate the total projection area of sand grains in the same particle size range and calculate the projection area ratio for the corresponding particle size range larger than.

[0108] S6: After calculating the projection area ratios for each particle size range larger than the particle size threshold, use three methods: the adjacent particle size range method, the average value method, and the weighted average method, to calculate the projection area ratio of small particle size sand grains in the global image and the local image respectively, that is, the projection area ratio for the small particle size range. This ratio can be used to obtain the total projection area of small particle size sand grains in the global image, where D t is the particle size threshold, R is the projection area ratio. After calculating the projection area ratios for the small particle size range deduced by the adjacent particle size range method, the average value method, and the weighted average method, calculate the variance for each method. The variance reflects the consistency and stability of the deduced results of each method. The smaller the variance, the more stable the result. By comparing the variances of these three methods, select the method with the smallest variance as the final calculation method to ensure that the deduced result of the projection area ratio for the small particle size range has the lowest error and the best stability, which can improve the overall calculation accuracy and ensure the reliability of the measurement results.

[0109] S7: Obtain the projection area ratio for the small particle size range. Map the small particle size sand grains in the local image to the global image to improve the area information of small particle size sand grains in the global image. Then, combined with the height H obtained in the single gradation shooting experiment, calculate the volume of sand grains in all particle size ranges in the global image, thereby calculating the gradation result of this group of sand grains.

[0110] For the same group of sand grains, after calculating the global gradation in different distribution states, by fusing the gradation data in these different states, the calculation accuracy and stability can be improved. By calculating the average value of the mass ratio for each particle size range in different states, the gradation results from multiple shootings are integrated into an overall result, which can effectively reduce the error caused by a single shooting and ensure that the final result is more accurate and reliable.

[0111] The preprocessing of the global image in S2 at least includes the following steps:

[0112] First, perform grayscale conversion and Gaussian filtering for denoising to retain the main contours. The standard deviation of the Gaussian filter is 2.

[0113] Use morphological "top-hat transformation" to highlight the bright areas, combine distance transformation to mark the particle centers, and cut the adhered particles through the watershed algorithm.

[0114] Finally, Otsu adaptive binarization is used to extract clear contours.

[0115] The preprocessing of the local image in S2 includes at least the following steps:

[0116] After grayscale processing, mild Gaussian filtering (standard deviation of 1) is used to retain details, with a standard deviation of 1;

[0117] Morphological "closing operation" is performed to fill small gaps, and the distance transform and watershed algorithm are combined to segment adhered particles;

[0118] Canny edge detection is used to extract the particle contours, and finally Otsu adaptive binarization is used to enhance details. The threshold of Canny edge detection is 50 - 150.

[0119] During the conversion in S3, due to different camera parameters (such as focal length, object distance) and image resolution, the actual pixel length of the sand grain size shown in the image is different.

[0120] Therefore, a pixel conversion formula is used to convert the actual particle size (in millimeters) into the corresponding pixel length in the image:

[0121]

[0122] where P represents the pixel value, f represents the camera focal length, fd represents the camera object distance, and D real represents the true size of the object.

[0123] The contour projection area calculation method used to calculate the contour projection area of each sand grain in S4 is the 45-degree outward expansion by half a pixel point method. The contour projection area calculation method effectively compensates for the influence of pixel discreteness and resolution on area estimation, thereby improving the accuracy of projection area calculation;

[0124] The specific operation of the contour projection area calculation method is to expand the contour of each sand grain along the 45-degree diagonal direction and calculate the pixel area occupied by the expanded contour, that is, to count the number of pixels in the filled area of the expanded contour, that is, the number of pixels in the white area, and use this as the contour projection area of the sand grain.

[0125] The steps of S5 are described as follows:

[0126] After obtaining the contour projection areas of all sand grains, calculate the total projection area of sand grains in the same particle size interval and calculate the projection area ratio of the corresponding particle size interval with a particle size greater than D t of

[0127]

[0128] where, is the total projected area of sand grains in a certain particle size range [D i , D i+1 in the global image,

[0129] and is the total projected area of sand grains in the corresponding particle size range in the local image.

[0130] Among them, the projected area of sand grains in the global image in the projected area ratio refers to the projected area of sand grains in this particle size range that includes the entire global image, and the contour information of sand grains in the local area is not separately removed;

[0131] When calculating the projected area ratio, although the values of the contour area and the short diameter calculated in different images are affected by factors such as shooting conditions and resolution, these influencing factors are fixed for specific sand grains. Therefore, the present invention chooses to uniformly incorporate these factors into the calculation of the projected area ratio without the need for additional conversion steps. This method can directly calculate the projected area ratio with the original contour parameters, thus avoiding the complexity brought by conversion and ensuring the accuracy of the final calculation of the projected area of small particle sizes in the global image.

[0132] The adjacent particle size range method in S6 is applicable to the case where the particle distribution is relatively uniform, in which the ratio of the projected areas of adjacent particle size ranges is usually relatively stable. Refer to the following formula:

[0133]

[0134] Among them: the particle size threshold is D t , and two particle size ranges [D t , D t+1 and [D t+1 , D t+2 close to this threshold are selected, and the projected area ratio of the small particle size range is estimated by the average value of their projected area ratios.

[0135] The average value method in S6 is applicable to the situation where the particle size distribution is relatively uniform. Because it assumes that the change of the projected area ratio of the particle size range is relatively smooth and there is no obvious uneven distribution, the projected area ratio of the small particle size range is estimated by calculating the average value of the projected area ratios of all known particle size ranges;

[0136] First, calculate the projected area ratio of each particle size range, and then average the ratios of the projected area ratios to obtain the average projected area ratio

[0137] Next, select two particle size ranges closest to the above average value, and estimate the projected area ratio of the small particle size range through the projected area ratios of these two ranges. Refer to the following formula:

[0138]

[0139] Wherein, n is the total number of particle size intervals, and t represents the position corresponding to the particle size threshold, which is used to determine the threshold division point of the particle size interval, so that the intervals after t and including t are regarded as the intervals larger than the particle size threshold.

[0140] Then, according to the average projected area ratio, select the two projected area ratios that are closest to it. At this time, the projected area ratio of the small particle size interval is calculated from the average value of these two projected area ratios.

[0141] The weighted average method in S6 calculates the projected area ratio of the small particle size interval by setting weighted coefficients according to the distribution of sand grains in different particle size intervals. The weighted average method assumes that intervals with a larger number of particles should be given higher weights, while intervals with a smaller number of particles should be given lower weights. The weighted coefficients are usually set based on the number and frequency distribution of particles in the particle size interval. Through weighted averaging, a more accurate projected area ratio can be obtained, especially suitable for sand grains with uneven particle size distributions. The formula is as follows:

[0142]

[0143] Wherein, represents the projected area ratio of the i-th particle size interval larger than the particle size interval. n is the total number of particle size intervals, and t represents the position corresponding to the particle size threshold, which is used to determine the threshold division point of the particle size interval, so that the intervals after t and including t are regarded as the intervals larger than the particle size threshold. α i is the corresponding weight coefficient. The weight coefficient is determined according to the number and frequency of sand grains in each particle size interval. The particle size interval with a larger number of particles corresponds to a higher weight coefficient. α i needs to satisfy:

[0144]

[0145] Example 2:

[0146] This example proposes a specific setting application based on the above Example 1;

[0147] Before implementing this method, it is necessary to pre-obtain the equivalent height of each particle size and the best calculation method for the projected area ratio of small particle sizes through multiple single-graded experiments and comprehensive-graded experiments. The single-graded experiment is because the quality of different sand and gravel materials is different, resulting in changes in their height H. Therefore, single-graded experiments are required to obtain corresponding effects. The comprehensive-graded experiment is because this method provides multiple calculation methods for the projected area of small particle sizes. For different situations, it is necessary to select the most suitable calculation method to ensure the accuracy of the final calculation result.

[0148] In image acquisition and preprocessing, the global sampling camera uses the model MVL-HF5024M-10MP, with a focal length set to 12 mm, an object distance of 178 mm, and the resolution of the captured images being 3072*2048. The local sampling camera is MVL-HF1224M-10MP, with a focal length of 50 mm, an object distance of 123 mm, and the resolution of the captured images being 5472*3648, which is specifically used to capture the details of sand grains with a particle size less than 0.3 mm. The vibrating disk is set to an intensity of 95 (full value 100) and a vibration frequency of 44 (full value 100) to ensure that the sand grains show different distributions in multiple captures. The actual sampling range is 107 mm x 23 mm for the global image and 33 mm x 23 mm for the local image. It is calculated that the area of the local image is approximately 10.75% of the global image.

[0149] First, preprocess the captured sand grain images. Since the sand grain information to be extracted is different for the global image and the local image, the present invention designs two image preprocessing methods.

[0150] For the global image, first perform grayscale conversion and Gaussian filtering (standard deviation of 2) to remove noise and retain the main contours. Use morphological "top-hat transformation" to highlight the bright areas, combine distance transformation to mark the particle centers, and cut the adhered particles through the watershed algorithm. Finally, perform Otsu adaptive binarization to extract clear contours.

[0151] For the local image, after grayscale processing, use mild Gaussian filtering (standard deviation of 1) to retain details. Perform morphological "closing operation" to fill small gaps, combine distance transformation and the watershed algorithm to segment the adhered particles. Use Canny edge detection (threshold 50-150) to extract the particle contours, and finally use Otsu adaptive binarization to enhance the details.

[0152] The global image processing focuses on stability and overall contour information, while the local image processing focuses on high precision and detail retention to avoid large errors in subsequent global gradation calculations.

[0153] To accurately divide the particle size range, first, it is necessary to convert the actual sand grain size into the pixel size in the image. Due to differences in camera parameters (such as focal length and object distance) and image resolution, the pixel length of the actual sand grain size in the image will vary. After calculation, the following are the pixel sizes of different particle sizes in the global and local images:

[0154] In the global image, the pixel sizes for 0.075 mm, 0.15 mm, 0.3 mm, 0.6 mm, 1.18 mm, 2.36 mm, and 4.75 mm are 133.43, 66.29, 33.15, 16.85, 8.43, 4.21, and 2.11 respectively. In the local image, the pixel sizes for 0.075 mm, 0.15 mm, 0.3 mm, 0.6 mm, 1.18 mm, 2.36 mm, and 4.75 mm are 804.54, 399.73, 199.86, 101.63, 50.81, 25.41, and 12.7 respectively.

[0155] Based on the above parameters, the global and local images can accurately divide the particle size intervals. Subsequently, the sand grain contour information within each particle size interval is obtained, and the projected area of the sand grains is calculated.

[0156] After calculating the projected areas of the sand grains in the particle size intervals greater than 0.15 mm in the global and local images, the corresponding projected area ratios are calculated. The projected area ratio of the small particle size is calculated by the optimal method for calculating the projected area of the small particle size, and then the volume of the sand grains is calculated by multiplying the projected area by the height, and the final grading result and fineness modulus difference are obtained.

[0157] Finally, by fusing the grading data in different distribution states, the accuracy and stability of the global grading calculation are improved, the accuracy of the final result is ensured, and the errors that may be brought by single-shot shooting are avoided.

[0158] In summary, the present invention solves the limitations of global and local images in particle measurement by introducing the particle size threshold, projected area ratio calculation, and equivalent volume characterization method, and successfully improves the measurement accuracy of fine particles. Through the dual-vision compensation algorithm, the present invention breaks through the contradiction between accuracy and sample quantity in traditional grading calculation on the premise of ensuring sufficient sample quantity, and improves the accuracy and stability of grading calculation. This method is applicable to the sand grain grading measurement in various industrial fields and has broad application prospects.

[0159] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, in any aspect, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claims involved.

Claims

1. A method for calculating the gradation of machine-made sand based on a dual-vision compensation algorithm, characterized in that: At least the following steps are included: S1: It uses dual cameras to collect global and local images of sand, and uses multiple vibration shooting sampling technology to obtain sand grain images. It takes multiple shots and samples for the same group of sand grains in different distribution states during the vibration process of the vibrating plate. The distribution and arrangement of the sand grains will change with each shot, which effectively reduces the errors that may occur in a single shot. S2: Preprocess the captured sand grain images. Two image preprocessing methods are used to extract different sand grain information from global and local images. Global image processing focuses on stability and overall contour information, while local image processing focuses on high precision and detail retention to avoid large error precision problems in subsequent global grading calculations. S3: After extracting the sand grain contour information after image preprocessing, it is necessary to convert the actual sand grain size into the pixel size in the image so as to accurately divide and classify the grain size intervals in the image; S4: After all the sand particles are divided into different particle sizes, the contour projection area of ​​each sand particle needs to be calculated so as to be used for the subsequent calculation of the projection area ratio of the particle size interval; S5: After finding the outline projection area of ​​all sand particles, calculate the total projection area of ​​sand particles in the same particle size range, and calculate the projection area ratio of the corresponding particle size range with a particle size greater than; S6: After calculating the projection area ratio of each particle size interval greater than the particle size threshold, the projection area ratio of small-size sand particles in the global image and the local image is calculated using the adjacent particle size interval method, the average method, and the weighted average method, i.e., the projection area ratio R of the small particle size interval. Dt , this ratio can be used to calculate the total projected area of ​​small-size sand particles in the global image, where D t is the particle size threshold, R is the projection area ratio. After calculating the projection area ratio of the small particle size interval estimated by the adjacent particle size interval method, the average value method and the weighted average method, the variance of each method is calculated respectively. The variance reflects the consistency and stability of the estimated results of each method. The smaller the variance, the more stable the result. By comparing the variances of the three methods, the method with the smallest variance is selected as the final calculation method to ensure that the estimated result of the projection area ratio of the small particle size interval has the lowest error and the best stability, which can improve the overall calculation accuracy and ensure the reliability of the measurement results. S7: Obtain the projection area ratio of the small particle size range The small-sized sand particles in the local image are mapped to the global image, and the area information of the small-sized sand particles in the global image is improved. Then, combined with the height H obtained in the single gradation shooting experiment, the volume of the sand particles in all particle size ranges in the global image is calculated, thereby calculating the gradation results of this group of sand particles.

2. The method for calculating the gradation of machine-made sand based on a dual-vision compensation algorithm according to claim 1, characterized in that: The preprocessing of the global image in S2 at least comprises the following steps: First, grayscale and Gaussian filtering are performed to remove noise, and the main contours are retained. The standard deviation of the Gaussian filtering is 2; The morphological "top hat transformation" is used to highlight the bright area, the particle center is marked by combining distance transformation, and the adherent particles are cut by the watershed algorithm; Finally, Otsu adaptive binarization is used to extract clear contours.

3. The method for calculating the gradation of machine-made sand based on a dual-vision compensation algorithm according to claim 1, characterized in that: The preprocessing of the local image in S2 at least comprises the following steps: After grayscale processing, a mild Gaussian filter (standard deviation is 1) is used to retain details, and the standard deviation is 1; Perform morphological "closing operation" to fill small gaps, and combine distance transformation and watershed algorithm to segment the adherent particles; Canny edge detection is used to extract particle contours, and finally Otsu adaptive binarization is used to enhance details. The threshold of the Canny edge detection is 50-150.

4. The method for calculating the gradation of machine-made sand based on a dual-vision compensation algorithm according to claim 1, characterized in that: When performing the conversion in S3, due to different camera parameters and image resolution, the actual sand grain size is displayed in different pixel lengths in the image. Therefore, the pixel conversion formula is used to convert the actual particle size into the corresponding pixel length in the image: Among them, P represents the pixel value, f represents the focal length of the camera, fd represents the object distance of the camera, and D real Indicates the true size of an object.

5. The method for calculating the gradation of machine-made sand based on the dual-vision compensation algorithm according to claim 4 is characterized in that: The contour projection area calculation method used in S4 to calculate the contour projection area of ​​each sand grain is the 45-degree outward expansion half pixel method. The contour projection area calculation method effectively compensates for the influence of pixel discreteness and resolution on area estimation, thereby improving the accuracy of projection area calculation; The specific operation of the contour projection area calculation method is to expand the contour of each sand grain along the 45-degree diagonal direction, and calculate the pixel area occupied by the expanded contour, that is, to count the number of pixels in the expanded contour filling area, that is, the number of pixels in the white area, and use this as the contour projection area of ​​the sand grain.

6. The method for calculating the gradation of machine-made sand based on a dual-vision compensation algorithm according to claim 5, characterized in that: The steps of S5 are described as follows: After finding the outline projection area of ​​all sand particles, calculate the total projection area of ​​sand particles in the same particle size range. And calculate the particle size greater than D t The projected area ratio of the corresponding particle size range in, is a particle size interval in the global image [D i ,D i+1 ] total projected area of ​​sand particles, It is the total projected area of ​​sand grains in the corresponding particle size range in the local image. The projection area of ​​the sand grains in the global image in the projection area ratio refers to the projection area of ​​the sand grains in the particle size range in the entire global image, and the sand grain contour information in the local area is not removed separately.

7. The method for calculating the gradation of machine-made sand based on the dual-vision compensation algorithm according to claim 6 is characterized in that: The adjacent particle size interval method in S6 is applicable to the case where the particle distribution is relatively uniform, wherein the ratio of the projected areas of adjacent particle size intervals is usually relatively stable, as shown in the following formula: Where: The particle size threshold is D t , select two particle size intervals close to the threshold [D t ,D t+1 ] and [D t+1 ,D t+2 ], and the projection area ratio of the small particle size range is estimated by the average value of their projection area ratio.

8. The method for calculating the gradation of machine-made sand based on a dual-vision compensation algorithm according to claim 6, characterized in that: The average value method in S6 is suitable for the case where the particle size distribution is relatively uniform, because it assumes that the projection area ratio of the particle size interval changes smoothly and there is no obvious uneven distribution. The projection area ratio of the small particle size interval is estimated by calculating the average value of the projection area ratio of all known particle size intervals; First, calculate the projected area ratio of each particle size range, and then average the projected area ratio to obtain the average projected area ratio. Next, select the two particle size intervals closest to the above average value, and use the projection area ratio of these two intervals to calculate the projection area ratio of the small particle size interval, refer to the following formula: Wherein, n is the total number of particle size intervals, t represents the position corresponding to the particle size threshold, and is used to determine the threshold dividing point of the particle size interval so that t and the intervals thereafter are regarded as intervals greater than the particle size threshold. Then, according to the average projection area ratio, select the two projection area ratios closest to it. At this time, the projection area ratio of the small particle size interval is Calculated as the average of the ratio of these two projected areas.

9. The method for calculating the gradation of machine-made sand based on a dual-vision compensation algorithm according to claim 6, characterized in that: The weighted average method in S6 calculates the projected area ratio of the small particle size interval by setting a weighting coefficient according to the distribution of sand particles in different particle size intervals. The weighted average method assumes that intervals with a large number of particles should be given a higher weight, while intervals with a small number of particles should be given a lower weight. The weighting coefficient is usually set based on the number and frequency distribution of particles in the particle size interval. Through weighted averaging, a more accurate projected area ratio can be obtained, which is particularly suitable for sand particles with uneven particle size distribution. The formula is as follows: in, represents the projection area ratio of the ith particle size interval that is larger than the particle size interval, n is the total number of particle size intervals, t represents the position corresponding to the particle size threshold, which is used to determine the threshold dividing point of the particle size interval, so that t and the intervals after it are regarded as intervals larger than the particle size threshold, α i is the corresponding weight coefficient, which is determined according to the number and frequency of sand particles in each particle size interval. The particle size interval with a larger number of particles corresponds to a higher weight coefficient. i Need to meet:

10. A method for calculating the gradation of machine-made sand based on a dual-vision compensation algorithm according to any one of claims 5 to 9, characterized in that: The S7 at least comprises the following steps: The total projected area of ​​small-size sand particles in the global image is defined as Total projected area of ​​small-size sand particles in the local image The total projected area of ​​sand particles in a certain size range in the global image is The total projected area of ​​sand particles in a certain size range in the local image is The projection area ratio is R, and the projection area ratio smaller than the particle size threshold is The projected area ratio of a certain particle size interval greater than the particle size threshold is R i , where i∈{t,…,n-1}, t represents the position corresponding to the particle size threshold, which is used to determine the threshold dividing point of the particle size interval; The total projected area of ​​small-size sand particles in the local image and projected area ratio The total projected area of ​​small-size sand particles in the global image is obtained by multiplication: is the particle size in the global image that is smaller than the particle size threshold D t The total projected area of ​​the sand particles. It should be noted that as long as the particle size is smaller than the threshold D t When calculating the total projected area of ​​these sand particles in the global image, the compensation coefficients are unified as Based on the fixed assumption of sand and gravel density ρ, using the formula Calculate the equivalent height H of specific graded particles; The mass M of sand and gravel was determined experimentally, and pictures of multiple groups of single-graded (i.e., single-size interval) sand grains were taken to calculate the total projected area S of each group of sand grains, and the formula Calculate the equivalent height h of each group of sand particles, then take the average of multiple groups of h to get the equivalent height H; The total projected area of ​​each group of sand particles is calculated by the forty-five degree outward expansion method of half a pixel point. Finally, a new sand and gravel equivalent volume characterization method was proposed: Sand grain projection area × equivalent height H The new sand and gravel equivalent volume characterization method is used for calculating the volume of each grain of sand, which is a batch of materials, that is, all sand grains with the same density. The sand grains are normally sampled and photographed. By combining the projected area of ​​the grains with the calculated equivalent height, the error caused by the irregularity of the grain shape is effectively overcome, and the accuracy of the volume calculation is significantly improved.

Images